Fixed Points and Random Fixed Points for Weakly Inward Approximable Maps Author(s): Donal O'Regan Reviewed

work(s): Source: Proceedings of the American Mathematical Society, Vol. 126, No. 10 (Oct., 1998), pp. 3045-3053 Published by: American Mathematical Society Stable URL: http://www.jstor.org/stable/119107 . Accessed: 03/02/2012 06:35
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Mathematical Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings of the American Mathematical Society.

http://www.jstor.org

OF THE PROCEEDINGS SOCIETY MATHEMATICAL AMERICAN Volume 126, Number 10, October 1998, Pages S 0002-9939(98)04601-2

3045-3053

FIXED POINTS AND RANDOM FIXED POINTS MAPS FOR WEAKLY INWARD APPROXIMABLE
DONAL O'REGAN (Communicated by Palle E. T. Jorgensen) In this paper we obtain new fixed point and random fixed point ABSTRACT. theory for approximable maps.

1. INTRODUCTION

This paper presents fixed point theory for weakly inward, approximable maps defined on a metrizable linear topological space (which is not necessarily locally convex). The paper is divided into two main parts. In Section 2 we use ideas in [1] (see also [9, 12]) to present new fixed point results for approximable maps. These maps were introduced in the Hausdorff linear topological space setting in [2, 3]. We will discuss in detail the setting when our map is also weakly inward; however other types of maps will also be considered. In Section 3 we use a recent result [14] to present random analogues of the results in Section 2. We now gather together some definitions and known facts which will be used throughout this paper. Let E be a Hausdorff linear topological vector space and let C be a closed, convex subset of E. The set Ic(x) = {x +?A(y-x) : A > O, y E C} for x E C is called the inward set at x. A mapping F: C -* 2E (here 2E denotes the family of all nonempty subsets of E) is said to be weakly inward with respect to C if F(x)nIc(x)

0 on C.

In fact, since C is closed and convex we have A> 1, y E C}. Ic(x) = x + {A(y-x): Let Y be a topological space. A mapping F: E -* 2Y is upper semicontinuous {x E E : F(x) n B $ 0} is closed for any closed set (u.s.c.) if the set F'(B) B in Y. Suppose X and Z are subsets of Hausdorff topological vector spaces E1 and E2 respectively and F: X -* 2Z is a multifunction. Given two open neighborhoods U and V of the origins in E1 and E2 respectively, a (U, V)-approximative continuous selection [2] of F is a continuous function s : X -* Z satisfying s(x) E (F[(x+ U) nX] + V) nZ for every x E X.

Received by the editors December 26, 1996 and, in revised form, March 17, 1997. 1991 Mathematics Subject Classification. Primary 47H04, 47H10, 47H40, 54C60, 54H25.
?)1998 3045 American Mathematical Society

3046

DONAL O'REGAN

F is said to be approximable [3] if its restriction FIK to any compact subset K of X admits a (U, V)-approximative continuous selection for every open neighborhoods U and V of the origins in E1 and E2 respectively. Let (Q, A) denote a measurable space. For a metric space (X, d) we denote by C(X) all nonempty closed subsets of X. Suppose Z is a nonempty subset of X and F: Z -+ 2X. Now F is called hemicompact if each sequence {(xn)}I 1 in Z has a convergent subsequence whenever d(xn, F(xn)) -> 0 as n -+ ox (recall that if Y is a nonempty subset of X then d(x, Y) = infyEy d(x, y)). A mapping F Q -> 2X is measurable (respectively weakly measurable) if for every closed (respectively open) subset D of X, F-1(D) = {w E Q : F(w) n L $& E A. A mapping 0} F :Qx Z 2X is called a random operator if for every x E Z, the map F(., x): X is measurable. A random operator F: Q x Z -+ 2x is called continuous (hemicompact, etc.) if for each w E Q, F(w, .) is continuous (hemicompact, etc.). The single valued map q : Q- X is said to be (i) a deterministic fixed point of F if q(w) E F(w, +(w)) for all w E Q and (ii) a random fixed point of F if q is a measurable map such that q(w) E F(w, +(w)) for all w E Q. Next we state a very recent result of Tan and Yuan [14].
* - *

Theorem 1.1. Let (Q, A) be a measurable space and Z a nonempty, separable, complete subset of a metric space (X, d). Suppose the map F : Q x Z -+ C(X) is a random, continuous, hemicompact map. Then F has a deterministic fixed point if F has a random fixed point. To conclude this introduction, we would like to state that the results of this paper will be used in forthcoming papers ("Fixed points, multivalued inequalities, control problems and differential inclusions on proximate retracts" and "Viable solutions of differential equations and inclusions on proximate retracts in Banach spaces") to establish new existence theory for viable solutions of differential inclusions on proximate retracts. 2.
FIXED POINTS

This section presents some fixed point theory for upper semicontinuous maps. Throughout this section E will be a metrizable (metric d) linear topological vector space (not necessarily locally convex), C will be a closed, convex subset of E, and Uo a d-bounded (i.e. there exists K > 0 with Uo C {x E E: d(O,x) < K}) open subset of E. We will let U = U0 n C and 0 E U. We consider maps F: U CC(E); here CC(E) denotes the family of all nonempty, compact subsets of E. Our discussion is restricted to the spaces IAPC (U, E), KAP(U, E), AP(U, E) and KIAP(U, E).
-

Definition 2.1. We say F E IAPC(U, E) if F: U -+ CC(E) is a u.s.c., bounded (i.e. F(U) is a subset of a d-bounded set in E), approximable, weakly inward w.r.t. C, map which satisfies property (B). Remark. Property (B) will usually mean that the map is compact. However in some situations (which we will specify at each stage) property (B) will mean that the map is condensing. We only introduce property (B) for convenience in presenting our results.

FIXED POINTS

FOR WEAKLY

INWARD

APPROXIMABLE

MAPS

3047

Definition 2.2. We say F E KAPc(U, E) if F: U map which satisfies property (B) and F(x) C Ic(x)

-+

CK(E) is a u.s.c., bounded

on U;

here CK(E) denotes the family of all nonempty, compact, acyclic [6, 7] subsets of E. Definition 2.3. We say F E APC(U, E) if F: U -+ CC(E) is a u.s.c., bounded, approximable map which satisfies property (B) and F(x) C Ic(x) on U. is a u.s.c., As-

Definition 2.4. We say F E KIAPC(U, E) if F: U -+ CK(E) bounded, weakly inward w.r.t. C, map which satisfies property (B).

Theorem 2.1. Let E, C, Uo5U be as above, 0 E U and G E IAPc(UE). sume the following conditions are satisfied: (2.1) if A C U satisfies A C co(G(A) U {0}),
{

then A is compact,

any map F E IAPQ(Q, E) has a fixedpoint; here Q is any d-bounded, closed, convex subset of E,
-+

and

(2.3)

J and any d-bounded, closed, convex set Q D U. 1 we have that N : Q - CC(E) (given below)
x N(x) = { z(x) G(x), E U. 0}, x EQ\U. 1X-

for any continuous map ,u: U

[0,1] with ,u(OU) =0

satisfies property (B); OU denotes the boundary of U -in C and here

Then either (Al) G has a fixed point in U; or (A2) there exist u e OU and A E (0,1) with u E AG(u). Remark. Suppose E is a Banach space, property (B) means the map is condensing and G : U -* CQ(E); here CQ(E) denotes the family of all nonempty, compact, convex subsets of E. That is, G E IAPc(U, E) if G : U -* CQ(E) is a u.s.c., bounded, condensing, weakly inward w.r.t. C, map (it is well known [3, 4] that G is approximable). In addition (2.1), (2.2) (see [5]) and (2.3) are clearly satisfied. We refer the reader also to [13]. Proof. Suppose (A2) does not hold and G does not have a fixed point on OU. Let H={xEU: xEAG(x) forsome AE[0,1]}. Now H 5 0 since 0 E H. A standard argument [10] implies H is closed. This together with (2.1) implies H is compact. Now since H n OU = 0 there is a continuous function ,u : U -* [0,1] with ,u(H) = 1 and ,a(OU) = 0. Since Uo is a d-bounded set and G E IAPc (U, E) (in particular G(U) is a subset of a d-bounded set in E), there exists R > 0 with

Uo C {x E E: d(O,x) < R}

and G(U) C {x E E: d(O,x) < R}.

3048

DONAL

O'REGAN

Let Q=C

n {x EE:

d(O,x) <R+1}

anddefine

r~x fjt(x) G(x), x EU, N(x) ={1 {}, x E Q!\U.

Now N: Q CC(E) is u.s.c., bounded, and satisfies property (B). Next we show that N is approximable. To see this, let K be any compact subset of Q. Our argument follows the technique in [2, page 486] (see also [12]). Let U1 and V/ be two open neighborhoods of the origin. We may assume without loss of generality that U1 is symmetric. Let V2 C V/ be a balanced open neighborhood of the origin such that V2 + V2 C V1. Since K is compact, G is u.s.c., and ,u is continuous, then for any x E K, it is easy to see (since G(K) is a subset of a bounded set) that there exists a neighborhood W, C U1 of the origin such that
-

,a([x + Wx)] n K) G([x + Wx)] n K) C ,a(x) G(x) + V2; here we use the notation ja(A) G(A) = { ja(x) G(y): x E A, y E A }. Let Zx C Wx be a neighborhood of the origin such that Zx + Zx C Wx. Now let {xi + Zxj}1 be an open cover of K and let U2 = nli Zxi Since K n U is compact, let s: K n U -* E be the (U2, V2)-approximative continuous selection of GIKnU*Let si(x) S,
=

u(x)s(x), x E Kn U, X E K n [Q\U].

We now check that si: K -* E is a (Ui, V1)-approximative continuous selection of NIK. Fix x E K n U. Then x E xi + Zxj for'some i E {1, ..., n}. Now since s(x)EG(x+U2)+ we have ,u(x) s(x) E ,a(xi + WxV) G(x + U2) + V2, and so ,A(x)s(x) E ,A(xi + Wxj G(xi + Wxj) + V2. Thus V2

,u(x) s (x) E ,ut (xi) + V1, G

and so n ,u(x) s(x) E ,u G ([x + U1fl K) + V1. Consequently si: K -* E is a (U1, Vl)-approximative continuous selection of We claim that N is weakly inward w.r.t. Q. If the claim is true then NIKN E IAPy(Q, E), and so by (2.2) we know that there exists x E Q with x E N(x). However, since 0 E U we have x E U. Thus x E N(x) and x E U. Consequently x E AG(x) with 0 < A = ,(x) < 1. As a result x E H, so ,u(x) = 1. Thus x E G(x), and we are finished. It remains to prove the claim. If x E Q\U then N(x) = {0} E IQ(x), since 0 e U0 n C (so 0 e Q). Next fix x E U. Now since G(x) nIc(x) then [s G(x) + (1 -s) 0] n Ic(x) since Ic(x) is convex. Consequently N(x) n Ic(x) 0 (where s =,u(x)) # 0 and 0 E IC(X),

$ 0.

FIXED POINTS

FOR WEAKLY

INWARD APPROXIMABLE

MAPS

3049

Now take y E N(x) = ,a(x)G(x) and y E Ic(x). Then there exist An > 1 for n E N, and {ZAn}fnEN C C, with d(yX+ Let
Vhn = X + An[ZAn
-

{An}nEN,

with

An[Z~n-x)

- 0 as n- oo.
X]

Now vvn
Let So
-

y as n
.

-*

00, so

VAn T-n=

E {x E E: d(O,x) < R + 1}
Then
Zin

for n E N sufficiently large.

= (1-

T) X + TnVAn,

Zin
E-

E Q for n sufficiently large. Thus

Zin

Q for n sufficiently large and d(y, x + An[zn

x])

-*

0 as n

00.

Thus y E IQ(x).

Remark. It is possible to replace Uo d-bounded and G(U) a subset of a d-bounded set in E by: there exists a convex set Co with Uo C Co and G(U) C Co [of course the other assumptions have to be appropriately adjusted]. A similar remark applies to the other theorems in this paper. Essentially the same reasoning as in Theorem 2.1 (except easier) establishes the corresponding results for KAP maps. Theorem 2.2. Let E, C, UO,U be as above, 0 E U and G E KAPc(UE). Suppose (2.1) holds. In addition assume that the conditions (2.4) and for any continuous map ,I: U
-*

any map F E KAPQ (Q, E) has a fixed point; here Q is any d-bounded, closed, convex subset of E
= [0,1] with ,aQ(JU) 0

(2.5)

and any d-bounded,closed, convexset Q D U.

we have that N: Q -* CK(E)

(given below)

satisfies property (B) are satisfied; here

01,

x E

Q\U.

Then either (Al) G has a fixed point in U; or (A2) there exist u E OU and A E (0, 1) with u E AG(u). Remark. Suppose E is a uniformly convex Banach space (or more generally a reflexive Banach space satisfying property (H) in [6]), and property (B) means the map is compact. That is, G E KAPC(U, E) if G : U -* CK(E) is a u.s.c., compact map with F(x) C IC(x) on U. Now (2.1), (2.4) (see [6, page 22]) and (2.5) are clearly satisfied. If E is a Hilbert space we could replace G being compact with G being condensing (see [6]). We refer the reader also to [8, 9]. Similarly we have results for AP and KIAP maps.

3050

DONAL O'REGAN

Theorem 2.3. Let E, C, UO,U be as above, 0 E U and G E APc(UE). pose (2.1) and (2.3) hold. In addition assume that the condition

Sup-

(2.6)
26Q

any map F E APQ(Q, E) has a fixed point; here is any d-bounded, closed, convex subset of E

is satisfied. Then either (Al) G has a fixed point in U; or (A2) there exist u E o9U and A E (0, 1) with u E AG(u). Theorem 2.4. Let E, C, Uo, U be as above, 0 E U and G E KIAPc(UE). Suppose (2.1) and (2.5) hold. In addition assume that the condition

(any map F E KIAP (Q,,E) has a

(2.7) fixed point; here Q is any d-bounded, closed, convex subset of E

is satisfied. Then either (Al) G has a fixed point in U; or (A2) there exist u E o9U and A E (0, 1) with u E A G(u). If E is Frechet and G is a condensing map, then (2.1) is automatically satisfied. However (2.1) may not be satisfied if G is a 1-set contractive map. It is of interest to try to extend Theorems 2.1-2.4 to this setting. We will just state and prove a result for IAP maps. A similar result holds for KAP, AP and KIAP maps (we choose to omit the details, since the reasoning is essentially the same in all cases). Theorem 2.5. Let E, C, U0, U be as above, 0 E U and G: U -* CC(E) be a u.s. c., bounded, approximable, weakly inward w.r. t. C, map. Assume the following conditions are satisfied: (2.8) for each n E {2,3, ...}, G, = (1--

G satisfies property (B);

(2.9)

['any map F E IAPQ (Q, E) has a fixed point; here Q is any d-bounded, closed, convex subset of E;

(2.10)

for each fixed n E {2,3, ...}, if A C U satisfies A C co(Gn(A) U {0}), then A is compact

(2.11)

I
and (2.12)

3,...}, and for any continuous map ,u: U -* [0,1] with ,a(OU) = 0 and any d-bounded, closed, convex set IQ D U, we have that N: Q)- CC(E) (given below) satisfies property (B) (if
{1X} C U with yn E G(xn) for all n and Xn - Yn - 0 as n -oo, then there exists x E U with x E G(x);

( for each fixed n E{2,

FIXED POINTS FOR WEAKLY INWARD APPROXIMABLE MAPS

3051

here
Nn(x)

{0},l xe ? \U.

{i (x)Gn(X),

XEU,

Then either (Al) G has a fixed point in U; or (A2) there exist u E o9U and A E (0,1) with u E AG(u). Proof. Suppose (A2) does not hold. For each n E {2, 3, ...} let Gn = (1) G. claim Gn E IAPc(U, E) for each n E {2,3,...}. We First we show that Gn is weakly inward w.r.t. C. To see this, fix x E U. Now since

G(x) n Ic(X) $70, 0 E IC(x)

and Gn(x)=
(i-k)

G(x) + 10, we have

Gn(x)

n Ic(x)

$7 0

since IC(x) is convex. Consequently Gn is weakly inward w.r.t. C. It remains to show that Gn is approximable. To see this let K be any compact subset of U. Let U, and Vi be two open neighborhoods of the origin. We may assume without loss of generality that Vi is balanced. Let s: K -* E be the (U1, V1)-approximative continuous selection of GJK, i.e. s(x) EG(x+U,)+Vi. Now clearly (1 -) s: K -* E is a (U1, Vi)-approximative continuous selection of (1- 1)GGK Consequently Gn E IAPc(U, E). We would like to apply Theorem 2.1. If there exist u E oU and A E (0,1) with u E AGn(u), then u E A (1 - n) G(u). This is a contradiction since (A2) does not hold. Apply Theorem 2.1 (with Gn replacing G), and we are guaranteed that Gn has a fixed point xn E U. Choose yn E G(xn) with xn = (1Yn. n) Notice that xn -Yn = - Yn - 0, since G(U) is bounded. Now apply (2.12) to deduce that there exists x E U with x E G(x). 1 3.
RANDOM FIXED POINTS

We now use Theorem 1.1 to produce random analogues of the results in Section 2. We prove a result for IAP maps (i.e., we will present the random analogue of Theorem 2.1). Similarly, random fixed point theorems coould be presented for KAP, AP and KIAP maps (i.e. we could establish the random analogue of Theorems 2.22.5); however we choose to omit the details since the reasoning is essentially the same in all cases. Throughout E will be a metrizable (metric d) linear topological vector space (not necessarily locally convex), C will be a closed, convex subset of E, and U0 a d-bounded open subset of E. We will let U = Uo n C and 0 E U. Theorem 3.1. Let E, C, Uo, U be as above and 0 E U. Also (Q, A) is a measurable space and U is a separable, complete subset of E. Assume G: Q x U -* CC(E) is a continuous, bounded, approximable, weakly inward w.r.t. C, hemicompact map which satsifies property (B) (this means that, for each w E Q, G(w, . )

3052

DONAL O'REGAN

is a continuous, bounded, approximable, weakly inward w.r.t. C, hemicompact map which satsifies property (B)). Suppose the following conditions are satisfied: (3.1) for each w E Q. if A C U satisfies A C co(G(w, A) U{O}) then A is compact; E) (for each w E Q, any map F(w, .) E IAPO (Qfo, has a fixed point; here Qo is any d-bounded, closed, convex subset of E;
r

(3.2)

(3.3)

for each w E Q, and for any continuous map p: U - [0, 1] with p(WU) = O and any d-bounded, closed, convex set QOD U. we have that N(w, . ): QO-* CC(E) (given below) satisfies property (B);

and (3.4) for any w E Q, u 4 A G(w, u) for all u E U and A E (O,1); here
A (x) N(w, x) = { {o}, G(w, x), x E U, x E Q\U. Then G has a random fixed point.

Proof. From Theorem 2.1, we know that G has a deterministic fixed point. Now O Theorem 1.1 implies G has a random fixed point.
REFERENCES 1. J. Banas and K. Goebel, Measures of noncompactness in Banach spaces, Marcel Dekker, New York, 1980. MR 82f:47066 2. H. Ben-El-Mechaiekh and P.Deguire, Approachability and fixed points for non-convex setvalued maps, Jour. Math. Anal. Apple, 170(1992), 477-500. MR 94a:54103 3. H. Ben-El-Mechaiekh and A.Idzik, A Leray-Schauder type theorem for approximable maps, Proc. Amer. Math. Soc., 122(1994), 105-109. MR 94k:54074 4. A. Cellina, A theorem on the approximation of compact multivalued mappings, Atti. Accad. Naz. Lincei Rend., 47(1969), 429-433. MR 43:2676 5. K. Deimling, Multivalued differential equations, Walter de Gruyter, Berlin, 1992. MR 94b:34026 6. P. M. Fitzpatrick and W. V. Petryshyn, Fixed point theorems for multivalued noncompact acyclic mappings, Pacific Jour. Math., 54(1974), 17-23. MR 53:8973 7. L. Gorniewicz, A. Granas and W. Kryszewski, Sur la methode de l'homotopie dans la theorie des points fixes pour les applications multivoques (partie 1: Transversalite topologique), C. R. Acad. Sci. Paris, Ser. 1, 307(1988), 489-492. MR 90g:55002 8. M. Lassonde, On the use of KKM multifunctions in fixed point theory and related topics, Jour. Math. Anal. Apple, 97(1983), 151-201. MR 84k:47049 9. D. O'Regan, Some fixed point theorems for concentrative mappings between locally convex linear topological spaces, Nonlinear Analysis, 27(1996), 1437-1446. MR 97i:47121 10. D. O'Regan, A topological approach to integral inclusions, Proc. Royal Irish Acad., 97A (1997), 101-111. 11. D. O'Regan, A continuation theory for weakly inward mappings, Glasgow Math. Journal, to appear. 12. S. Park, Fixed point for approximable maps, Proc. Amer. Math. Soc., 124(1996), 3109-3114. MR 96m:47108

FIXED

POINTS

FOR WEAKLY

INWARD

APPROXIMABLE

MAPS

3053

13. S. Reich, A fixed point theorem in Frechet spaces, Jour. Math. Anal. Apple, 78(1980), 33-35. MR 82h:47055 14. K. K. Tan and X. Z. Yuan, Random fixed point theorems and approximation in cones, Jour. Math. Anal. Apple 185(1994), 378-390. MR 95d:47085 ,
DEPARTMENT OF MATHEMATICS, UNIVERSITY COLLEGE GALWAY, GALWAY, IRELAND

E-mail address: donal. oreganlucg. ie